Karaman, F. . & Hay, J.
The University of Tennessee, Knoxville
In this study, we examine infants' long-term memory for statistically-defined words found in natural language. In Experiment 1, English-learning 8-month-old infants (n=20) were familiarized with a set of Italian sentences that contained two words with high transitional probability (TP =1.0) and two words with low TP (TP = .33). Discrimination of HTP from LTP word was tested using Headturn Preference Procedure immediately following familiarization (T1), and following a 10-minute delay (T2). Although infants failed to discriminate HTP from LTP words when tested immediately after familiarization, the same infants looked significantly longer to HTP words (mean=10.2 sec) than LTP words (mean= 9.1 sec) after the 10-minute delay, t(19)= 2.41, p < .05. To investigate the potential influence of exposure to isolated words during the first testing session (T1), in Experiment 2 infants (n=25) were familiarized to the corpus and were then tested after a 10-minute delay (T2 only). Without the additional support of experience with isolated words, infants in Experiment 2 failed to demonstrate discrimination of HTP from LTP words. We will discuss the possibility that instead of interfering with word-internal probabilities, isolated words may selectively reinforce long-term memory for sound sequences with high internal transitional probabilities.